(
single_estimator_variance: f64,
average_correlation: f64,
n_estimators: usize,
)
| 205 | } |
| 206 | |
| 207 | pub fn bagging_ensemble_variance( |
| 208 | single_estimator_variance: f64, |
| 209 | average_correlation: f64, |
| 210 | n_estimators: usize, |
| 211 | ) -> Result<f64, String> { |
| 212 | if single_estimator_variance < 0.0 { |
| 213 | return Err("single_estimator_variance must be non-negative".to_string()); |
| 214 | } |
| 215 | if !(-1.0..=1.0).contains(&average_correlation) { |
| 216 | return Err("average_correlation must be in [-1,1]".to_string()); |
| 217 | } |
| 218 | if n_estimators == 0 { |
| 219 | return Err("n_estimators must be > 0".to_string()); |
| 220 | } |
| 221 | |
| 222 | let n = n_estimators as f64; |
| 223 | let rho = average_correlation; |
| 224 | Ok(single_estimator_variance * (rho + (1.0 - rho) / n)) |
| 225 | } |
| 226 | |
| 227 | pub fn recommend_bagging_vs_boosting( |
| 228 | base_estimator_accuracy: f64, |
no outgoing calls